[ https://issues.apache.org/jira/browse/FLINK-24743 ]
Martijn Visser deleted comment on FLINK-24743:
----------------------------------------
was (Author: martijnvisser):
Twice more on release-1.20 with the AdaptiveScheduler, both in
test_file_sink.sh (New File Sink and Streaming File Sink s3):
https://dev.azure.com/apache-flink/98463496-1af2-4620-8eab-a2ecc1a2e6fe/_build/results?buildId=80014&view=logs&j=fb37c667-81b7-5c22-dd91-846535e99a97
(nightly 2026-10-10)
https://github.com/apache/flink/actions/runs/39356761314/job/118431519304
(nightly 2026-10-20)
After the second kill the job never sees the second restart, and the
ResourceManager keeps sending slot requests to the TaskManager that was killed:
{code}
Killing 2 TMs
...
Still waiting for restarts. Expected: 2 Current: 1
...
2026-10-20 03:58:06,344 WARN
org.apache.flink.runtime.resourcemanager.slotmanager.DefaultSlotStatusSyncer []
- Slot allocation for allocation b7196bf416cfd79013a0983ceaaa0c99 for job
da7c5b486324c85f91b709cf82c78f6f failed.
org.apache.flink.runtime.rpc.exceptions.RecipientUnreachableException: Could
not send message [RemoteRpcInvocation(TaskExecutorGateway.requestSlot(...))] ...
Test (pid: 122883) did not finish after 900 seconds.
{code}
> New File Sink end-to-end test fails on Azure
> --------------------------------------------
>
> Key: FLINK-24743
> URL: https://issues.apache.org/jira/browse/FLINK-24743
> Project: Flink
> Issue Type: Bug
> Components: Connectors / FileSystem, Runtime / Coordination, Tests
> Affects Versions: 1.15.0, 1.20.6
> Reporter: Till Rohrmann
> Priority: Major
> Labels: auto-deprioritized-major, test-stability
>
> The {{New File Sink end-to-end test}} fails on Azure. It seems that the
> {{TaskExecutor}} cannot connect with the {{ResourceManager}} because
> {{Discarding inbound message to
> [Actor[akka://flink/user/rpc/taskmanager_0#1155592017]] in read-only
> association to [akka.ssl.tcp://flink@localhost:6123]. If this happens often
> you may consider using akka.remote.use-passive-connections=off or use Artery
> TCP.}}
> https://dev.azure.com/apache-flink/apache-flink/_build/results?buildId=25831&view=logs&j=c88eea3b-64a0-564d-0031-9fdcd7b8abee&t=070ff179-953e-5bda-71fa-d6599415701c&l=11674
> cc [~chesnay], this might be related to the recent Akka bump.
> Related: https://github.com/akka/akka/issues/24393
> Maybe we should upgrade to using Artery TCP or set
> {{akka.remote.use-passive-connections=off}}.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
